testing-workflow

Run the Makefile targets mapped to each tier: unit client tests use jsdom with Testing Library, Apollo server tests run under node, integration enforces global 100% coverage, and Playwright handles E2E flows. All invocations go through the profile's make target map rather than bare jest or playwright commands.

Review the mutation diff in the output, add a test case covering the uncaught mutation, and strengthen assertion specificity. For example, if a mutant changes > to >=, add a boundary test case. Never lower the MSI threshold or silence mutants with suppression annotations.

Open the diff image and decide whether the change is intended. Unintended diffs are regressions requiring a component or style fix; intended changes allow regenerating the baseline snapshot only after inspection. Never update snapshots blindly to silence a failing run.

No, this workflow prohibits adding data-testid attributes. Use accessible role, label, and text locators in both Testing Library and Playwright, refactoring to semantic queries when an element is hard to find.

Snapshots drift when test data is non-deterministic. Use seeded Faker builders via the shared tests/builders helpers, with the default seed or FAKER_SEED override, so generated data stays reproducible and baselines remain stable across runs.

Skip a suite only when its capability flag is false or its mapped make target is null, and record an explicit capability-absent note. Otherwise the mutation gate enforces the configured MSI threshold and visual tests enforce zero unintended diffs.